MIT Paper: AI Agent Autonomously Conducts 18.9-Hour Quantum Experiment
imjustnewatai · x · 2026-07-30
A new MIT preprint details an 18.9-hour autonomous quantum sensing experiment where an AI agent selected quantum defects in diamond, calibrated resonance, and designed pulse sequences, while deterministic software retained physical control for safety.
More Reasoning Leads to More Hallucinations
Evaluating GPT-5.4 to 5.6, researchers found a counterintuitive trend: when given only pulse sequence context, higher reasoning effort made models more likely to hallucinate non-existent signals. GPT-5.5's false-positive rate surged from 14.8% at low reasoning to 53.2% at xhigh.
A Design Rule for AI Scientists
The fix requires the model to calculate an expected signal quantitatively before making a judgment, holding false positives to 0–3.7% across all settings. The authors propose a core design principle: let AI explore and hypothesize, but force claims through quantitative predictions and keep physical control deterministic.
Related event: MIT Showcases AI Agent Running 18.9-Hour Quantum Experiment(2 posts)→
More from coding & agent
- Evaluating Agents Without Right Answers: Similarweb's Playbook — LangChain · 2026-07-30
- AIPOCH Open-Sources Library of 550+ Medical Research Agent Skills — tom_doerr · 2026-07-30
- O'Reilly Author Proposes: Replacing Hardcoded Agent Workflows with Natural Language — JnBrymn · 2026-07-30
- Multi-Agent Coding Tested: The Orchestrator Must Know When to Disappear — RFOK · 2026-07-30
- hf-mount-encrypted: Mount HF Buckets with Client-Side Encryption — jedisct1 · 2026-07-30
- Indie Dev Showcase: Building a Multi-Account Ad Dashboard with Convex — TJLarkin23 · 2026-07-30